Evolutionary approaches for pooling classifier ensembles: Performance evaluation

Abstract

We introduce a multiple classifier system that incorporates an Evolutionary Algorithm for dynamically selecting the set of classifiers to be included in the pool. The proposed technique is applicable when the classifiers provide both the class assigned to the input sample and a measure of thereliability of the classification. For each sample, the experts selected for participating in the voting rule are those whose reliability is larger than a given threshold. There are as many thresholds as the number of classifiers by the number of classes. The problem of finding the values of the thresholds aimed at selecting the best set of classifier for each input sample has been reformulated as an optimization task, approached by using the Breeder Genetic Algorithm and the Differential Evolution. A set of experiments on three well-known and widely adopetd datasets have been designed and performed to compare the performance provided by the two competing approaches.